Is the market still pricing JiZhijia (02590) according to “hardware equipment vendors”? It's already doing “physical AI”

The 2026 World Artificial Intelligence Conference (WAIC) has just come to an end, and the “humanoid robot show” at various booths is still impressive — Yu Shu's humanoid robot G1 performed difficult backflips, while Zhiyuan Expedition A2 showed smooth human-computer interaction.

However, one company's booth had a completely different style of painting.

Ji Zhi Jia (02590) doesn't dance, doesn't walk, doesn't talk. At the booth, the Gino 1 wheeled robot, unmanned picking workstation, and several mobile robots are cooperating with high accuracy to quietly pick potato chips, bread, and rubber bowls one by one from the bin and accurately place them in the corresponding planting wall position. There is no extra “performance”; only a complete set of end-to-end work processes is running stably — sorting, handling, and delivery. Every step is as smooth and silent as in a real warehouse or factory.

The action isn't dazzling — but what the industrial scene wants isn't dazzling; it's stability. Among the more than 150 smart exhibitors of WAIC this year, Gizhijia is the only one that simultaneously demonstrates multi-SKU generalization capabilities and end-to-end operation efficiency in real work scenarios.

According to the Zhitong Finance App, Jizhijia's core ambition is not just to create better robots, but to be a “solution expert” in the intelligent era — based on more than ten years of accumulation of real scenarios, to provide a thousand industries with embodied intelligent solutions that can be implemented, replicated, and scaled. This is a strategic transition from “champion products” to “expert services”, and it is also a core clue to understand the long-term value of Gizhijia.

Gravity Framework: “An Intelligent Operating System”

In 2006, when Amazon AWS was quietly launched, no one thought it was a big deal. At the time, everyone's eyes were on Google and YouTube — dazzling apps that really “changed the world.” Looking back twenty years later, the market value of AWS surpassed the sum of almost all the star apps of that era.

Today, the wave of physical intelligence is sweeping through at an alarming rate. A large number of companies are keen to show cutting-edge humanoid robot demos, but JiZhijia did the opposite, showing the stable performance of its wheeled robot Gino 1 in complex, multi-SKU, and generalized scenarios, with “real machine hard work” as the core.

Behind this “bland” is a profound strategic judgment: the future of embodying intelligence does not depend on the brilliance of individual robots, but on whether a unified, reliable, and open “operating system” can be built.

Jizhijia's core release of WAIC this time is a set of unified embodied intelligence frameworks called Gravity.

Gravity uses a “dual brain collaboration” architecture: the cognitive brain is responsible for understanding instructions and disassembling tasks; the action brain performs physical “sandbox deduction” before executing actions to generate optimal action sequences after predicting consequences. Chen Chao, head of the Jizhijia Sized Model Team, made a statement on the opening day of the WAIC, accurately pointing out the core contradiction in the current industry: “Models that understand language semantics are not accurate physical motion, and models that can predict physical motion lack high-level semantic understanding.”

What is more critical is Gravity 4D's technical route selection. It breaks out of the limitations of “only reproducing images”, enhances 3D space and 4D physical representation of motion, takes into account reasoning speed and cross-material generalization ability, and uses full-body parallel collaborative control to adjust the grabbing posture simultaneously, increasing single-task efficiency by about 30%.

Data verification: The success rate of zero samples in the Libero-Plus benchmark increased from 73.73% to 78.62%, and the three “appearance becomes physically unchanged” scenarios, such as changes in camera viewing angle, sensor noise, and lighting changes, improved most significantly.

GINO ECO Ecosystem: Builder and Evangelist of “Standards”

The iteration of technology necessarily requires ecological support.

The “one core and two engine” strategy launched by Jizhijia at this time closely integrates the Gravity framework (one core) with the data engine and ecosystem engine (dual engine), outlining its platform-based ambitions.

If OpenAI's barrier is Internet text data, Gizhijia's barrier is interactive data in the physical world and an open ecological engine.

This is not a gap that can be quickly caught up with financing and recruiting people. The real business network that Jizhijia has accumulated over 11 years — covering more than 40 countries, over 1,700 projects, and serving more than 950 internationally renowned brands (Walmart, Adidas, Siemens, BMW, etc.) — generates millions of SKU real data capture data every day, forming an almost insurmountable data moat, and this moat is still expanding at an accelerated pace.

This time, Jizhijia simultaneously launched the GINO ECO open ecosystem: open the hardware platform for algorithm companies, explore innovative scenarios with leading customers, and create an end-to-end workflow with complementary hardware companies.

Many people understand this as a “win-win gesture”. But to understand its deep logic, we need to look at another historical case: Android. By opening up Android to all mobile phone manufacturers, Google ostensibly relinquished hardware control; in fact, it used openness in exchange for access control of the mobile internet, which ultimately led to the prosperity of the entire Android ecosystem.

Gizhijia's logic is the same. The more open GINO ECO is, the more algorithm companies and hardware companies that connect, and the more difficult it is to shake the Gravity framework's status as an “interface standard”.

As Zheng Yong, founder and CEO of Jizhijia, said, “Embodying intelligence is a long-distance race, not a sprint. We want to be a superconnector for the entire industry.” By connecting all links, Jizhijia is constructing an embodying intelligent panorama “with Gravity as the core, data as fuel, and ecology as leaves”.

Poor perception is a real pricing opportunity

Behind the strategic ambition, it is supported by business logic that has already worked.

In fiscal year 2025, JiZhijia handed over a report card that crossed the inflection point of profit: revenue of 3.171 billion yuan, an increase of 31.6% over the previous year; adjusted net profit of 43.82 million yuan, which turned a loss into a profit; and a sharp correction in operating cash flow. This is a company that has crossed the inflection point of profit and is not still burning money to tell a story.

Meanwhile, management has repurchased 16.2 million shares over 15 consecutive trading days since June 25 — the signal speaks for itself.

The explosion of the smart market has become the consensus of the industry. Goldman Sachs Research predicts that the global humanoid robot market will reach 38 billion US dollars by 2035; Citi is more optimistic. It expects the global humanoid robot market space to reach 154 billion US dollars by 2035.

Thanks to the general trend of the industry and the company's fundamentals, out of the 8 brokerage firms covered by Jizhijia, all gave a buy or increase rating. Among them, CITIC Securities had a maximum target price of HK$53 and an average comprehensive target price of about HK$24.6 — there is room for an increase of more than 120% compared to the current stock price.

But more important than the spread is a shift in cognitive frameworks.

What the market is currently offering is the valuation framework for “Warehouse AMR Hardware Company”. But what Gizhijia is becoming is a “physical AI operating system” company. The valuation gap between these two frameworks is not 20% or 30% room for repair, but may involve a fundamental reorganization of pricing logic — as happened back then when the market repriced NVIDIA from “gaming graphics card companies” to “AI computing infrastructure.”